The causal effect of relocation

Estimates of the average time-varying effects of relocation on the hazard rate into first job are shown in Table 6.3. The effects of the first move and of more than one move are estimated separately for 6 different intervals of time since the move: 1-3 months, 4-6 months, 7-12 months, 13-18 months, 19-24 months and more than 24 months. For men, relocation away from the assigned municipality on average has a significantly positive effect on the hazard rate into first job and the effect increases with time since the move. On average leaving the assigned municipality also has a significantly positive effect on the hazard rate into first job for women, except in the first time interval in which the effect is close to zero. The time-varying effect of moving twice relative to staying in the assigned municipality is also significantly positive in most time-intervals and increasing with time since the second move.

Table 6.3 Estimated effects of relocation on the hazard rate into first job.

Sub sample: Men Women

Variable: Coefficient Std. error Coefficient Std. error

One move:

Move 0-3 months ago 0.216* 0.103 -0.016 0.215

Move 4-6 months ago 0.520** 0.091 0.568** 0.164

Move 7-12 months ago 0.489** 0.073 0.384** 0.139

Move 13-18 months ago 0.466** 0.078 0.589** 0.134 Move 19-24 months ago 0.464** 0.083 0.391** 0.153 Move > 24 months ago 0.408** 0.055 0.614** 0.088 Two moves:

Second move 0-3 months ago 0.374* 0.165 -0.018 0.447 Second move 4-6 months ago 0.270 0.181 0.814* 0.334 Second move 7-12 months ago 0.311* 0.136 0.212 0.319 Second move 13-18 months ago 0.564** 0.131 0.585* 0.293 Second move 19-24 months ago 0.504** 0.141 0.525 0.307 Second move > 24 months ago 0.568** 0.084 0.595** 0.154

Notes: One and two asterisks indicate significance of the estimate at the 5 and 1 % levels, respectively. Controls for demographic and socio-economic characteris-tics and time-invariant unobserved characterischaracteris-tics of the individual and observed location characteristics are included in the employment model. In addition, selec-tion into relocaselec-tion is taken into account.

For men, relative to no move the hazard rate into the first job is increased by 24% 1-3 months after the first move, 68% 4-6 months after the first move and around 58% from then on. For women, relative to no move the hazard rate into first job is increased by 0% 1-3 months after the first move, 76% 4-6 months after the first move and 47%-85% from then on. The average effects of having moved twice are slightly higher.

These positive effects of carrying out a cross-municipal move are likely to stem at least in part from differences in unobserved location characteristics before and after the move, such as presence of an ethnic network in the municipality of destination which can facilitate job search in the new local labour market. Not only may these unobserved factors affect the local job offer arrival rate, they may also contribute to lower the reservation wage by increasing place utility (and thus increase the local job offer acceptance rate).

Since we only observe when an individual begins in a job and not - which would have been preferable - when an individual actually got the job, one may

worry that the positive effects at least in part stem from reverse causality;

some movers may have found a job in another region prior to moving to the region. However, if that were the case, we would expect the hazard rate into first job to be very high immediately after relocation and to decline thereafter.

Our estimates of the time-varying effect of relocation on the hazard rate into first job show exactly the opposite. For both sexes the two relocation effects are estimated to increase over time since the move. Furthermore, the first relocation effect for men is estimated to be small 1-3 months after the move.

Similarly both relocation effects for women are estimated to be zero 1-3 months after the move. Furthermore, descriptive evidence shows that out of the 3,720 individuals who find their first job outside the local labour market in the observation period only 248 relocate to another municipality within a six months interval around the time of beginning in the first job. This evidence does not support the hypothesis that refugees first search for a job outside commuting distance and then move once they have accepted a distant job. Therefore, we contend that the large positive effect of relocation on the hazard rate into first job is in fact a causal effect.

Besides the ’direct’ relocation effects reported in Table 6.3, relocation af-fects the hazard rate into first job through changes in the observed location characteristics, i.e. if observed characteristics of the municipality of destina-tion differ from those of the municipality of origin. We refer to this relocadestina-tion effect as the relocation effect due to observables. We calculate the average relocation effect due to observables in the following way,

(

X^locationdenotes the mean of observed municipality characteristics across movers, ’before’ refers to the beginning of the first residential spell and

’after’ refers to the time immediately after relocation. Table 6.4 reports the relocation effect due to observables for the first move and for two moves. On average, this effect is negative, and for men it is statistically significant. For women, the effects are insignificant at the 5% level. Hence, (male) refugee migrants moved to locations with less favourable employment prospects than in the assigned municipality. Specifically, the average effect of changes in location characteristics after the first move corresponds to a decline in the hazard rate into first job by 12-13%. The average effect of changes in location characteristics after two moves is larger; it corresponds to a decline in the

hazard rate into first job by 19% for men and 15% for women. The difference in the size of the relocation effect due to observables after one move and two moves is explained by the fact that relocations from a small or medium-sized municipality to a large municipality constitute a larger share of second-time relocations than first-time relocations.

Table 6.4 The relocation effect due to observables on the hazard rate into first job.

Sub sample: No. of moves Relocation effect due to observables Effect Std. error

MEN One -0.134 0.061

Two -0.212 0.068

WOMEN One -0.146 0.085

Two -0.167 0.088

The total average effect of relocation on the hazard rate into first job is the sum of the direct effect of relocation and the effect due to observables.

Hence, relative to no move the hazard rate into first job for men is increased by 8% 1-3 months after the first move, 46% 4-6 months after the first move and around 38% from then on. Relative to no move the hazard rate into first job for women decreased by 15% in the first three months after the first move, but 4-6 months after the first move it is increased by 51% and from 7 months after the first move and onwards it fluctuates between an increase of 26%

and 59%. The total average effect of having moved twice can be calculated similarly. Thus, in most cases the hazard rate into first job declines initially after relocation, but a few months after relocation it increases to a level which significantly exceeds the hazard rate in the case of no relocation.

Are relocation effects of the estimated size also economically significant?

We can answer this question by calculating the effect of relocation on the predicted mean duration of first non-employment spell. The predicted mean duration of first non-employment spell for a person with observed character-istics X and unobserved charactercharacter-istics v^ˆu is given by

E(Tu|X,v^ˆu) = Table 6.5 summarizes the estimated relocation effects in the following way: first we calculate the predicted mean duration for an individual who

does not move. Next, we calculate the change in the predicted duration for three different relocation scenarios: one relocation 16 months after initial settlement, two relocations in months 16 and 31, and finally one relocation in month 37. The latter simulation is meant to reflect the effect of the main new feature in the current Danish dispersal policy relative to the first Danish dispersal policy, namely that placed refugees are not entitled to social benefits in the municipality of destination if they leave the assigned municipality in the first three years after assignment.⁶ For an average woman the relocation effects correspond to a decrease in the estimated time until first job of 41 months (corresponding to a 34% reduction) if she moves 16 months after initial settlement and of 39 months if she moves twice. For men, the effects are 22 months for one relocation (corresponding to a 27% reduction) and 24 months for two relocations. All effects are statistically significant. Thus, only the first relocation really helped jobless refugee immigrants getting into their first job.

Table 6.5 Predicted mean duration and change in mean duration of first non-employment spell for an individual with average characteristics.

Predicted Predicted change in mean duration mean duration

Number of moves: None One Two One

Sub sample: Time of move: Month 16 Months 16, 31 Month 37

MEN Type 1 (v^ˆu= 0) 58.5 (1.86) -14.4 (1.37) -15.4 (1.55) -9.4 (0.91) Type 2 (v^ˆu= -1.018) 135.1 (8.00) -38.9 (3.87) -46.2 (4.54) -33.7 (3.41) Average person 82.6 (2.84) -21.9 (2.06) -24.4 (2.51) -16.8 (1.62) WOMEN Type 1 (v^ˆu= 0) 97.0 (5.32) -30.5 (3.55) -29.1 (5.35) -24.6 (2.77) Type 2 (v^ˆu= -0.920) 206.2 (21.92) -77.4 (7.97) -74.2 (13.60) -70.8 (6.77) Average person 122.6 (6.73) -41.3 (4.27) -39.3 (7.54) -34.9 (3.82) Notes: All durations are measured in months. An average male person is

0.45·type1+0.55·type2. An average female person is0.571·type1+0.429·type2^. Standard errors calculated by Monte Carlo methods are reported in parentheses.

The introduction of a three year relocation restriction under the current Danish dispersal policy implies a cost in terms of employment. For an average

6Obviously, such an interpretation should be made very cautiously, since the change in rules may affect the estimated parameters of the model.

individual who in the absence of the relocation restriction move away 16 months after initial settlement, the introduction of a three year relocation restriction is predicted to increase time until first job by 5.1 months for an average man and by 6.4 months for an average woman.

The results discussed above support the hypothesis that refugees moving improved the employment prospects of individuals. We have shown that, de-spite the fact that observed municipal characteristics are on average ’better’

in the initial municipality, movers still gain from moving, which supports the hypothesis of a reservation wage effect; individuals postpone job search until they have found a satisfactory municipality in which to reside, and then they lower their reservation wage in order to find employment. An extension of the introduction period from 18 months to 3 years would lead to a lower reduction in expected time until first job for those who move.